The valuation of environmental externalities : a stated preference case study on traffic noise in Lisbon

This research study was motivated by the need to valuing environmental externalities from road transport. The main objective was to develop a methodology centered on the Stated Preference-choice (SP-choice) method for valuing traffic noise when individuals are in their homes. The aim was to assess the nature and extent of households’ heterogeneity of preferences for quiet. For this purpose, two different metrics of the noise variable were used to estimate the marginal values of quiet, ratings based on household’s perceptions and the physical noise measures in Leq dB(A).
An innovative computer survey model was developed and administered to more than 400 households in a residential area in Lisbon with high-rise residential buildings in the vicinity of mam roads. The experimental design explored respondents’ familiarity and experience with perceived noise levels indoors in various apartment situations and at different floors of the block (lot). Lower and upper floors and their exposure to road traffic (fronting the main
road or located at the back façade) played a central role in the experimental design. A range of situational, socio-economic, behavioral and attitudinal variables relating to each household were collected. Physical noise measurements were taken at each apartment (indoors and at the exterior façade) and related to respondents’ perceptions. Complementary methods such as the revealed preference (RP) data on apartment purchases and the openended contingent valuation method (CVM) were also included. The issue of convergent validity of noise value estimates for the same sample of respondents was explored.
Multinomial Logit models including additional effects (MNL-INT) of a wide range of variables were explored, as well as combined MNL-INT with additional variables with
random parameters’ logit specifications (Mixed Logit, ML). In brief, the study found that models based on respondent’s perceptions outperformed those based on physical noise
measures. A range of other influential variables were found to interact with householders’ preferences such as adjusted household income per person, sign of noise changes (improvements or deterioration in the levels), floor number, base noise level experienced, and others. The ML specifications gave a better fit with the data. The income elasticity of marginal values of quiet was of similar magnitude in the SP-choice and RP methods, but a
weak income effect was detected when using the CVM data. Nevertheless, the strategic bias may have affected both the SP-choice and CVM experimental markets. The noise value
estimates were in the range of estimates found in other studies.